Chapter 4
The detection of metal-bending action
It occurred to me at a very early stage, during the Society for Psychical Research 1975 experiments carried out with Graham and Valerie P. at City University, London, that very small elastic deformations might be of common occurrence in metal-bending sessions; they would be undetectable by eye, and would result in no permanent deformation; but they could possibly be detected by instruments. Therefore I thought it would be useful to provide the metal specimens with a sensitive device capable of detecting small strains (elongations or deformations), such as are produced by, or at any rate associated with, stress (force or moment). I found the resistive strain gauge(22) to be by far the most suitable instrument for this purpose. One modern type of gauge consists of a thin film, typically of the alloy constantan, in the form of a folded filament, deposited on a small thin plastic sheet (see Figure 4.l). The gauge is either stuck onto a metal surface or onto the inner surface of a cavity machined in the metal; the plastic is sufficiently thin (<= 0.05 mm) for it not to reinforce the metal but to bend or extend with it. The electrical resistance of the constantan filament varies proportionally to its length (normal length plus strain), so that a suitable battery operated electrical bridge circuit and amplifier (Figure 4.1, as first built by my technician Nick Nicola) produces a time-varying voltage proportional to time-varying strain. This 'analogue signal' can be recorded on magnetic tape, or better still on a chart-recorder; thereby we obtain a measure of how long the strain lasts and how large it is. Suitable design enables very small strains to be measured, so that weak 'powers' of a paranormal metal-bender can be detected even though the specimen is far below the condition of permanent deformation. Calibration of the instrument is summarized in Table 4.1.
Table 4.1
Strain sensitivity delta l/l = epsilon = 3.33 X 10^-6/mV
Moment sensitivity on Al strip 12 mm wide, 0.75 mm thick sigma l = 2 X 10^-4 Nm/mV
Temperature sensitivity of the thermistor delta T = 2 x 10^-3 °C/mV
Standard deviation of sensitivity for six channels, for identical stress, ± 0.03.

Figure 4.1 Bridge and amplifier circuit for use with resistive strain gauge and chart-recorder; strain gauge (Micro-Measurements type EAO9 125 BT 120) is shown enlarged.
I built the device with electrical screening of the wires, sufficient to ensure that artefacts would not occur from electrostatic or electromagnetic causes. In the initial version a gauge was stuck onto a brass strip by Mr Chapman at the City University, and the boy whom we were studying (Graham P.) was allowed to stroke the strip with the ball of his forefinger. To our surprise some sharp little pulses appeared on the chart-record. When I myself stroked the sensor in a similar manner no pulses appeared. But at this stage Chapman and I were not convinced that we had recorded the phenomenon for the first time. We were both sceptical and thought that electrostatic artefacts must have been responsible. The signals were small, and I thought it possible that they could have arisen from the finger's jumping from point to point, in the manner of chalk producing an involuntary dotted line on a blackboard. In retrospect it now seems possible that these were genuine paranormal signals, but I decided at the time to make two drastic changes:
1 The strain gauge was to be enclosed within a cavity in the metal, and electrically screened as far as possible.
2 The child must not be allowed to touch the metal with his fingers.
With the new policy l started exposing strain gauges in latchkeys to the children l visited, Belinda H., Julie Knowles, Andrew G., Willie G., Richard B., Nicholas Williams, Alison Lloyd, Mark Henry, David and Steven Nemeth, Clifford White, Gill Costin and Neil Howarth. All of them managed to produce signals without touch, and l have since successfully exposed the sensors to other subjects in other countries.
Children enjoyed playing with the latchkeys in which the first strain gauges were mounted (Plate 4.1). When the key is deformed elastically between the fingers and thumb, a deflection can be seen on the chart-record, with the paper moving typically at a speed of 1 inch every 5 seconds. A latchkey is difficult to deform permanently by finger pressure alone, so the chart-recorder trace returns to approximately its original level. The device is such a pleasant toy that we might recommend its production commercially. Of course we must avoid the use of ferromagnetic metals, since spontaneous relaxation of domains might cause signals due to the Barkhausen effect.
The moment of truth for a metal-bender is when he reaises that similar signals appear when neither the fingers nor indeed any part of the body touches the latchkey. Sometimes I would place the latchkey on a plastic dish or in a glass bowl, and allow the metal-bender to stroke the glass beneath it. The best part of an hour was often necessary to produce the first paranormal signal, and it was almost as though the experimenter were coaxing it out of the metal-bender. The viewing of the chart-recorder is a form of biofeedback, and is best when the sensitivity is raised until the electrical noise shows as 'grass' on the trace; a tiny artefact is as much encouragement as the appearance of a tiny paranormal signal, and it does not matter if at first the two are confused. Larger signals are then to be expected. If they appeared when the key was resting in a glass bowl, I would then remove the bowl and allow the key to hang by its own electrical connections. The child would then just sit facing the key, possibly pointing his forefinger, or even stroking his fingers and thumb together.
Plate 4.1 Latchkey with milled slot containing resistive strain gauge. Epoxy-resin potting and screened lead, as used in early experiments with Nicholas Williams
Of course the apparatus must be allowed to run 'quiet' for as long as possible (sometimes hours) before exposure to the metal-bender. No signal must appear when the metal-bender is not present. Battery operation is preferable, and precautions must be taken against inductively coupled mains artefacts and atmospheric artefacts (e.g. those arising from strong walkie-talkie radio sets). Dummy resistances, circuits and chart-records are used for this purpose, and any signal appearing also in the dummy channel must be rejected.
The drift experienced on the strain gauge amplifier, although by no means large after the settling period, is not such that slow variations of strain could reliably be studied with the direct current system. The use of an oscillatory input, with or without phase-locked loop, would readily allow of this investigation, but we have postponed it for the time being.
The latchkey is suspended from its wires so as to minimize mechanical coupling; the child can point his fingers at the metal, at about six inches distance. We must watch carefully to see that there is no touching, but we should not appear to do so. If signals are still forthcoming, then the distance can be slowly increased. Most of the successful metal-benders have succeeded at distances of up to about three feet; Andrew G. and Stephen North have succeeded at up to eight feet and Nicholas Williams has regularly worked at fifteen feet, and once obtained a signal at thirty feet. Occasionally the latchkey or piece of metal actually bends permanently during its suspension from the strain gauge wires.
Figure 4.2a-e shows the entire chart-record of the first really impressive strain gauge session, which resulted in a permanent deformation of the latchkey. Nicholas Williams was seated by my side on the sofa in the lounge (see Figure 5.1) whilst the key was hanging by its wire from the mantelpiece on the opposite wall. We had a long wait before signals appeared, and I gave Nicholas cutlery which I asked him to bend. He twirled the pieces around, and in the course of time several displayed a 'curly bend'. My object in allowing him to handle cutlery was to encourage his possible 'power' to blossom and spill over onto the strain gauge specimen. Also his hands were occupied, so that any question of tampering with the strain gauge specimen was ruled out. During the session I changed the sensitivity of the chart-recorder amplifier at appropriate times. I shall never forget my increasing excitement as the small, doubtful signals increased in size, although I was completely confident of the reliability of my equipment, since prolonged tests had been made in Nicholas's absence, and there were no signals. Eventually the latchkey developed a permanent bend.
A most important lesson which is taught to the metal-bender by use of the strain gauge detector is the avoidance of over-concentration; he must learn 'sudden inattention' if he wants results. The insomniac who concentrates his mind on going to sleep will have little success; it is often the same with paranormal metal-bending. The strongest events often take place, it appears, when the subject relaxes directly after concentration. He must learn to be patient and not to 'concentrate on concentration'. If the experimenter makes it obvious that he is scrutinizing or is suspicious of the subject, then the effects are likely to be weak or altogether absent; if the experimenter conceals his scrutiny, and the metal-bender avoids over-concentration, then some signals may come. Parapsychologist Julian Isaacs has embarked on a series of experiments with groups of subjects and an audio-recorder, studying the psychological tensions at moments when signals are recorded.
Figure 4.2a-e Chart-record of dynamic strain gauge signals produced at the first session with Nicholas Williams. Vertical time scale markings of 5 sec are interrupted during quiet periods (e.g. 6 min, 4.5 min). Full-scale sensitivities are 100 mV, 1O mV, 100 mV again, 1 V and 10 V; square-ended signals indicate full scale.
Of course this 'learning of inattention', on the part of all present, is particularly unfortunate from the point of view of the observer. On the one hand he must develop reliable techniques of observation and instrumentation; on the other hand he must learn to appear inattentive if he wishes the effects to be stronger. Obviously some compromise is necessary. Most of the time is spent keeping a careful watch; but if rapport between the experimenter and the child is sufficiently good, then there is less danger of touching if the head is turned. One must judge whether the child is himself really interested in the experiment; in that case he will report an inadvertent touch. But always the experimenter must return to the pattern of good video observation. If bad habits of touching should develop, we can make use of the touch detector discussed further in chapter 15; we can also use partial screening of the target, and even a moving target.
When the child and the family first reaise that bending can take place without the necessity of the metal being touched, this is a great step forward in their understanding. Sometimes the child is a little frightened by this realization, and sometimes he never faces up to it at all.
If we are to make the detection of strain within untouched metal a valuable method of investigation, then we must encourage its use by other investigators. At the time of writing, ten other groups of workers have successfully used essentially the same equipment as ours; their reportings of signals obtained with various subjects can be regarded as an important confirmation of our findings. The most extensive experiments have been carried out in France and in Japan.
I have several times made video-records of strain gauge experiments, notably with Julie Knowles and Stephen North; also I have made a point of asking physicists and others to be present at the production of strain signals in the homes of these families. On occasion I have been asked to provide a metal-bender to produce signals on a television programme. Four times Stephen North has attempted this, and twice he was successful, twice he was not; but the effects cannot usually be produced to order, particularly when there are television lights, continual delays, photographic tests, makeup, and so on. I had the impression that Stephen was acting the part of a metal-bender rather than being his normal self. Certainly, Uri Geller, JeanPierre Girard and Masuaki Kiyota have learned sufficient control to work under these conditions. But there are many reasons why the metal-bending children should not be encouraged to follow in their footsteps.
An important choice the experimenter must make is the size of the metal specimen and just how the strain gauge should be mounted in it. When a strip of metal is gently stretched, the extension (strain) which causes the signal is directly proportional to the tension (stress). The thinner and narrower the strip, the larger is the extension for the same tension and the more sensitive is the equipment to force. When we require the greatest sensitivity, we may stick the gauge onto a metal strip only 0.75 mm thick and cover it with a thin piece of foil for electrostatic shielding. But the experiments concerned with the details of strong effects, such as those with Nicholas Williams or JeanPierre Girard, require much more robust specimens.
We must of course use the recommended strain gauge adhesives; there is a danger of paranormal forces tearing away the strain gauge from the metal (chapter 9). To attach the strain gauge with adhesive tape (a reported procedure) is quite inadequate. If the strain gauge itself is fractured by paranormal action, then the experiment terminates; but I have conducted sessions in which there was fracture of the specimen without destruction of the strain gauge and it was still possible to continue work.
When the strain pulses are sufficiently strong for the specimen to be deformed permanently, the chart-recorder trace may also show a permanent deflection. There might be a correlation between the magnitude of this deflection and the observed angle of the bend; an effort has been made to reproduce the data graphically in Figure 4.3 for sensors mounted in latchkeys in the Nicholas Williams experiments.(23) But the correlation is not an accurate proportionality, partly because the bend does not always take place exactly at the position of the strain gauge and partly because, as we shall find in chapter 6, the signal may correspond only to a permanent extension and not to permanent bend. A latchkey will usually bend somewhere along the part that goes into the lock, while the strain gauge has been deliberately mounted in the handle and will continue to function even when the latchkey is fractured. In the first session with Stephen North, using a long thin strip of aluminium, a very large visible bend or curl through 540° was produced more than an inch from the strain gauge, and no permanent deflection of the chart-record was manifest.
Figure 4.3 Comparison of permanent deflection chart-records with observed bend angles of latchkey during the Nicholas Williams sessions summarized in Table 5.1. The 45° straight line is merely a fitting to the data.
Sometimes the electrical connections to a strain gauge are bent or fractured paranormally during a session; in some early experiments it was my practice to embed them in solid epoxy-resin; but it is sometimes important to avoid alteration of the mechanical properties of the specimen.
When the strain gauge is embedded in the metal, the subject has no direct knowledge of its form, and it is very likely that any paranormal action will be on the metal itself. We have found evidence from subsequent experiments with several sensors mounted inside one specimen that it is usually the metal and not the strain gauge which experiences the 'action'.
During the first ten hours' experience with strain gauges exposed to metal-benders I learned much about their use from mistakes that were made. But after nearly two hundred hours of exposure l consider that I know sufficient to avoid mistaking drifts or artefacts for paranormal signals, and vice versa. Among others, physicists Ron Miller, David Robertson and Elizabeth Rauscher have taken part in the exposures of strain gauges and have brought a fresh critical approach to the methods l use.
It is of course important to verify by other experiments that the signals do not arise from paranormal action on the electrical equipment or on the pen of the chart-recorder, or simply from electric mains or electromagnetic artefacts coupled inductively to the battery-operated equipment. Two subsidiary experimental programmes were mounted for this purpose. In the first a galvanometer mirror was mounted on a very thin spring steel strip, with strain gauge attached. One end of the spring was attached to a stable horizontal surface under a glass dome, and an optical beam from a helium-neon laser was passed through the dome and was reflected through it again from the mirror and onto a scale. The overall optical path was about 6 m. Small movements of the light spot were seen to synchronise with strain gauge signals, and some ringing was observed, due to the long-period mechanical resonance in the system. One spring exposed to Stephen North became permanently deformed. This is the experiment which seems in retrospect to have been a miniature of the Bealings Bells situation (chapter 1).
In the second type of experiment a dummy strain gauge is included with real strain gauges on a metal strip. Signals are observed on the real strain gauge chart-record throughout a session, but signals are not recorded from the dummy gauge. Sometimes a stable resistor mounted on or near the metal is used as a dummy strain gauge. Sometimes a resistive thermal sensor (Micro-Measurements type STG 50D) of 70 Ohms resistance is connected with compensation in one channel of the electronics. In physical appearance the thermal sensor is similar to the resistive strain gauge (Micro-Measurements type EA09 125 BT 120), but its resistance is insensitive to strain although highly sensitive to temperature. Sudden temperature changes are unknown to us in paranormal metal-bending sessions, although temperature drifts usually occur, arising from convection currents (chapter 14). Paranormal strain signals on the other hand are sudden, in that they are sharp-fronted pulses; the measured durations are now known to be, on occasion, less than a millisecond.
The use of a chart-recorded dummy strain gauge channel is now standard practice in our experiments. Both types of experiments have vindicated the interpretation that paranormal action develops an internal strain in the metal or the strain gauge or both.
Naturally I have given much thought to the question of whether l consider the many thousands of strain pulses recorded in these sessions to be merely artefacts. It is quite possible to devise electronic methods of simulating signals, using electromagnetic coupling, but I am releasing no detailed information about how this might be done! The dummy strain gauge would show up such simulations, as it shows up artefacts arising from local electrical disturbances. Any signal also occurring in the dummy strain gauge channel (and such an event is rare) is discarded. There is also the possibility of someone jogging the chart-recorder mechanically or tampering with the electronics; obviously I am aware of this possibility, but I have never known it happen except accidentally. I think it possible that very small electromagnetic artefacts (about 0.2 mV in magnitude) might appear on the strain gauges but not in the dummy channel. But the vast majority of paranormal signals are much larger than this, and l do not see how they can represent anything other than paranormal strains.
Action on 'nude' strain gauges which are not stuck onto pieces of metal was observed at several no-touch sessions with Nicholas Williams. Typically the strain gauge was stretched between two rigid mounting points, and in some experiments a small piece of insulated wire was allowed to rest freely on it. The purpose was to search for any downward quasi-force exerted during bending of the wire. In fact such quasi-forces were observed each time the wire bent visibly. Their magnitudes sometimes exceeded those of the force exerted by the entire weight of the wire resting on the centre of the horizontal strain gauge. Thus these quasi-forces appear not to be simply changes in the weight of the metal; the interpretation of this type of experiment is at present rather difficult.
l believe it likely that the action was not on the strain gauge but on the wire, which bent. But in later experiments a strain gauge unloaded with wire was found to experience signals; possibly these were electrical in origin (chapter 15).
I have successfully exposed strain gauges mounted on many varieties of material; tungsten, brass, aluminium, copper, silicone rubber, wood, plastics, glass and fused silica.
With the strain gauge I have attempted to measure the strains experienced by a metal specimen during no-touch bending action. The action occurs in occasional bursts. There are characteristic patterns which these data exhibit. First, the signal rise times; are they sudden or gradual in their onset and their termination? Inspection of the data shows that nearly always the strain pulses have sharp onsets and usually sharp terminations. The time-resolving power of the chart-recorder (about 0.l sec) is in fact the limit of the observed sharpness, but other methods of recording have been used. It is rare for a more gradual onset of force to be recorded. When signals are a thousand times more powerful than typically, the rise times are slower, partly because the chart-recorder amplifier is slower. But typical signals show onsets which are sharper than those which can be recorded from finger action on the metal; finger signals, except for the smallest, are softened by muscular response and flesh distortion.
I have amplified the paranormal signals and listened to them acoustically; they are bumps rather than clicks; i.e. frequencies greater than about 500 Hz appear to be insignificant. Occasionally there is acoustic ringing of the metal specimen, and occasionally the metal specimen is seen to swing a little on its wire suspension.
Second, the duration of the signals; although most are of less than a second, a few are of several seconds' duration. Occasionally the pulse (Figure 4.4a) appears to be continuous with wobbles superposed. Possibly this represents an unresolved group of shorter signals. There are some signal events in which a pulse in one direction is followed immediately or after one or two seconds by a pulse in the other direction (Fig. 4.4b). Activity continuing for more than about ten seconds is rare; although sometimes there is minor activity ('nibbling', Figure 4.4c) which can occasionally last for more than a minute. Sometimes there is extensive structuring of the signals (Figure 4.4d). Sometimes the elastic component of the strain is suppressed, so that a bend is achieved without over-swing (Figure 4.4e). Sometimes an elastic pulse in one direction is followed by a plastic deformation in the other (Figure 4.4f).
Third, the magnitude of the signals. Considered as stress signals they depend upon the strength of the specimen on which the strain gauge is mounted. Results of calibration experiments of the system in terms of strain and stress are given in Table 4.1. Signals during a metal-bending session can be distributed over several orders of magnitude, from a few millivolts to a few volts on the chart-recorder, so that it is difficult to make precise predictions. Both lower and upper limits are instrumental, the former being the electrical noise, and the latter the upper limit of the chart-recorder (10 V) and sometimes the fracture of the sensor or of the metal. It is usual for the signals to increase from small to large during a session; only occasionally are the first signals the largest. Sometimes there is a continuous train of signals of similar magnitude, either small or large. There are many factors, psychological and physical, which determine the signal magnitude, and they will be discussed further in chapter 17.
Figure 4.4 Some dynamic strain pulses showing distinctive features. All are from Nicholas Williams sessions B-F. Time scale intervals are all 5 sec. Two-channel measurements (chapter 5), with sensors numbered 1 and 2, in separate latchkeys.
a, signals with reasonably flat tops; b, signal followed immediately by one in the opposite direction; c, nibbling; d, structured signals; e, permanent deformations in which the elastic components are suppressed; f, elastic strain pulses in one direction followed by plastic strain pulses in the opposite direction.
Fourth, the structure of the signal events in time, which we have already illustrated and described in Figure 4.4. Examination of even this small amount of Nicholas Williams data will give some idea of the great variety and unpredictability of the structure. Such records should be rare in purely physical experiments, where the structure should show at least some regularity; the signal records are strongly reminiscent of biophysical or environmental data, which, of course, they are. Any attempt to explain their origin must take account of this variety of fine structure. Some of it cannot be produced by hand. There is some discernible variation between the patterns of signal as between different metal-benders. Most signals produced by metal-benders are simply sharp peaks, but a few are followed by reverse peaks, or overswings, which are not instrumental in origin. The reverse signals are in this case just as sharp as the parent signals preceding them.
An interesting type of signal makes its appearance in Figure 4.4 (3 EH f). It has a curved tail, always of the same form, an 'exponential' exp(). The characteristic time varies only with the design of the sensor; it is invariant during a session, for example. Perhaps these tails arise from physical relaxation of internal stress in which both the metal, the strain gauge and the adhesive play their part (see chapter 9).
During the reception of signals, I have studied the no-touch movement of a specimen and strain gauge suspended from its electrical connections; this has been done visually, and on several occasions by means of a video-camera. I noticed that when a relatively strong dynamic strain pulse was produced by Nicholas Williams the key would sometimes be joggled a few millimetres and would swing for perhaps a second. This motion was first recorded on videotape with the cooperation of Dr Ron Miller. But the magnitude of the movements was small, and we cannot make any further generalizations from a study of the tapes, other than that the key was not touched by any visible agency. A stop-clock was included in the field of view, in order that the small movements of the key could be synchronised with the signals recorded on the chart-recorder.
Subsequently David Robertson and I made video-records of strain gauge experiments with Stephen North and Julie Knowles. When a long, thin (0.75 mm X 8 mm X 40 cm) metal specimen suspended at each end received paranormal strains at the centre, it sometimes continued its bending oscillations, which were recorded until they died out in a few seconds due to physical damping.
A feature of the sessions with both Stephen North and Nicholas Williams was the slow start with gradually increasing frequency of occurrence of signals. It is possible that they were learning during the session to lose their inhibitions against producing signals. Several children have commenced sessions with small signals, separated by long periods of time; but not all of them make the transition into the more powerful regime, when signals come thick and fast.
In the next chapter we proceed to the discussion of the information that can be obtained by the use of more than one detection device at a time.

Back to books list.
Back to main index.
Back to Uri Geller's home page.